31/3/2016

Beyond the Hashtags: Asking data how #Blacklivesmatter impacted offline conversations

 

In the summer of 2014, the United States' public agenda was dominated by the issue of police brutality. On Twitter, #Blacklivesmatter garnered over 52,000 mentions, and social posts were often pinpointed as drivers behind offline conversations on the topic. To test the impact of this online activism, Deen Freelon, Charlton D. McIlwain, and Meredith D. Clark, conducted an analysis using three kinds of data:

1) Twitter data

Since social tools are usually employed by activists at pivotal junctures in the movement's development, the team divided over 40.8 million relevant tweets into time periods corresponding to defining events:

blmtable.PNG

Then, they conducted a content and network analysis at each juncture to find isolate the top users, what they were saying, and their place within the broader anti-brutality movement. In doing so, the researchers were able to document interactions between Twitter happenings and the wider public consciousness.

"First we separated the authors and full text of all tweets into the nine time periods described above. Next we created a network edge list for each period that connected usernames (nodes) to one another on the basis of retweets or mentions (edges), so that each period was represented by its own network. This was an effective means of analyzing this data given that well over 80% of the tweets consisted of retweets and mentions," they explain.

"We then generated a set of network communities within each period’s edge list using an algorithm called the Louvain method. The Louvain method creates communities by maximizing edge density within communities and minimizing it between communities. For very large networks like ours, Louvain creates small numbers of very large communities and large numbers of very small communities (many of which consist of a single user retweeting or mentioning another once). In each period, we analyzed only the 10 largest communities, which in most cases accounted for well over half of all users involved in retweets or mentions."

To analyze this Twitter data, the team used a Python-based module called Twitter Subgraph Manipulator.

2) Web link data

To ensure a comprehensive understanding of the online movement, the Twitter analysis was complemented by a broader online network analysis. Through a paid subscription to the hyperline analysis tool VOSON, the team identified 100 sites as the basis for their crawl. These sources were chosen based on their known affliations with the #Blacklivesmatter movement, as well as Google rankings.

From this crawl, the team discovered 136,587 links to and from the original 100 sites. These links were then grouped by site to paint a picture of the main media in the conversation and the networks connecting these. Ultimately, 1,243 separate sites were identified, with 5,115 links between them.

The following visualization, for example, illustrates the vast connections to and from the Blacklivesmatter.com domain.

blmcom.PNG

In addition to the basic link data, metadata, such as visitor demographics and traffic, was also collected using Alexa.com, UrlProfiler, and Buzzstream’s MetaTag Extractor. This data allowed the researchers to understand who was engaging with conversations on these sites. For example, this visualization highlights the overrepresentation of key demographics access networked sites:

demographicsblm.PNG

3) Qualitative interview data

To add context to the twitter and hyperlink network analysis, the researchers collected and analyzed qualitative data from 40 interviews.

"The interviews were analyzed for descriptive codes which were used to identify participant thoughts, feelings and motivations as expressed in the conversations. Next, analytic codes were developed through an iterative process of reviewing the data, and were used to create categories for analysis."

The results

By combining their results from the twitter, network and qualitative data, the team made the following observations about online and offline conversations about police brutality:

  • Although the #Blacklivesmatter hashtag was created in July 2013, it was rarely used through the summer of 2014 and did not come to signify a movement until the months after the Ferguson protests.
  • Social media posts by activists were essential in initially spreading Michael Brown’s story nationally.
  • Protesters and their supporters were generally able to circulate their own narratives without relying on mainstream news outlets.
  • There are six major communities that consistently discussed police brutality on Twitter in 2014 and 2015: Black Lives Matter, Anonymous/Bipartisan Report, Black Entertainers, Conservatives, Mainstream News, and Young Black Twitter.
  • The vast majority of the communities observed supported justice for the victims and decisively denounced police brutality.
  • Black youth discussed police brutality frequently, but in ways that differed substantially from how activists discussed it.
  • Evidence that activists succeeded in educating casual observers came in two main forms: expressions of awe and disbelief at the violent police reactions to the Ferguson protests, and conservative admissions of police brutality in the Eric Garner and Walter Scott cases.
  • The primary goals of social media use among interviewees were education, amplification of marginalized voices, and structural police reform.

Read the full research report here.

Image: Joe Brusky

Comments